This action might not be possible to undo. Are you sure you want to continue?
A Panel Data Analysis of Electricity Demand in Pakistan Azam Amjad Chaudhry∗
Abstract This paper looks at the economy-wide demand and the firm level demand for electricity in Pakistan. The economy wide estimation of electricity demand uses panel data from 63 countries from 1998-2008, and finds that the elasticity of demand for electricity with respect to per capita income is approximately 0.69, which implies that a 1% increase in per capita income will lead to a 0.69% increase in the demand for electricity. The firm level analysis uses firm level data from the World Bank’s Enterprise Survey for Pakistan and finds that the price elasticity of demand for electricity across all firms is approximately -0.57, which implies that a 1% increase in electricity prices will lead to a 0.57% decrease in electricity demand across firms. Across sectors, the textile sector has the highest price elasticity of demand (-0.81) while the price elasticity of demand for firms in the electricity and electronics sector is the smallest (-0.31). Finally, firm level data is also used to estimate production functions in order to estimate the impact of electricity shortages on manufacturing output. Keywords: Electricity, demand, industrial, price elasticity, Pakistan. JEL Classification: Q41, E39, E01. I. Introduction In a period in Pakistan’s economic history when the amount of electricity generated has fallen far short of the economy-wide demand, an analysis of electricity demand is critical both for understanding why the country has reached this point and for determining future policy measures to tackle this shortage. Although the topic of electricity demand in Pakistan is vast, this paper looks at the economy-wide demand for electricity and also focuses on industrial-level demand. The questions that this paper attempts to answer are questions that the average Pakistani, the Pakistani business community, and Pakistani policymakers are attempting to answer: What will be the economy-wide
Associate Professor and Dean, Department of Economics, Lahore School of Economics.
Azam Amjad Chaudhry
demand for electricity in the coming years? What impact will higher electricity prices have on industrial demand for electricity? How does electricity affect industrial level output? Each question is a topic on which vast research is possible, but this paper looks at each using two unique datasets. Most analyses that try to determine future electricity demand in Pakistan focus on time series data for Pakistan. This paper takes a different approach: Instead of just focusing on Pakistan, why not use historical data to see how electricity demand has grown over time in other countries? Using a unique panel dataset for 66 countries over 10 years taken from the International Energy Agency (IEA) and the Penn World Tables, the relationship between income per capita and electricity consumption per capita is estimated. This is then used to predict the growth in electricity demand in Pakistan over the next 5-10 years. The second set of analyses using data taken from the World Bank Enterprise Surveys for Pakistan to construct a panel dataset for 402 firms over 4 years. This dataset is used to determine the elasticity of electricity demand with respect to the price of electricity across firms in Pakistan. This can be used to gauge the impact of industrial electricity price increases on the demand for electricity in the Pakistani industrial sector. This dataset is also used to estimate the production function for Pakistani firms; this estimation uses a production function that not only includes the usual factors of production, labor, and capital, but also includes electricity. With these results, one can estimate the impact of reductions in electricity supplies on industrial output. The setup of the paper is as follows: Section II provides a brief background on electricity production and prices in Pakistan, and compares it with other countries. Section III uses the IEA dataset to estimate the elasticity of electricity consumption per capita with respect to real income per capita. Section IV uses the World Bank Enterprise Survey dataset to estimate the demand for electricity in the Pakistani manufacturing sector. Section V uses the World Bank Enterprise Survey dataset to estimate a firm level production function with electricity. Section VI contains conclusions. II. Background In 1947, Pakistan inherited 60 MW of power generating capacity for a population of 31.5 million people and this was expanded to 119 MW by 1959, just as the country was entering a phase of development that required reliable infrastructure. In the interim, the government acquired a majority shareholding of the Karachi Electric Supply Company (KESC) in 1952 and created the Water
A Panel Data Analysis of Electricity Demand in Pakistan
and Power Development Authority (WAPDA) in 1958, the purpose of which was to coordinate the development of schemes in water and power. After its first 5 years of operation, WAPDA expanded the electricity generation capacity to 636 MW and increased the number of electrified villages from 609 to 1,882. This expansion picked up speed in the 1970s and 1980s with the generating capacity rising to 1,331 MW in 1970, followed by further growth to 3,000 MW by 1980 and 7,000 MW in 1990/91. However, rapid urbanization, industrialization, and rural electrification initiatives led to a significant growth in the demand for electricity (at a rate of 9–10% per annum) which meant that peak electricity demand fell short of supply by the early 1990s, requiring approximately 1,500–2,000 MW of load shedding. This shortfall was dealt with by the government with the ‘Policy Framework of Package of Incentives for Private Sector Power Generation Projects’ in 1994, leading to private sector initiatives in power development and an expansion of the supply to the point at which there was surplus power. But it should be noted that the policy led to the dramatic expansion of thermal projects and resulted in the reversal of the hydel/thermal power generation mix. In 2000, the restructuring of WAPDA started, which eventually resulted in WAPDA being broken down into 14 separate units: Four thermal generating companies, nine distribution companies, and a transmission and distribution company. The main electric power producers in Pakistan are WAPDA, KESC, the Pakistan Atomic Energy Commission and a number of independent power producers (IPPs) established since 1994. Table-1 shows the current situation in the electricity sector, with an estimated shortfall of 4,000 MW in 2008/09 and an estimated shortfall of 5,500 MW in 2009/10. Table-2 and Figure-1 show the breakdown of electricity supply in Pakistan and also illustrate the reversal of the hydel/ thermal power generation mix after the 1994 package. Figures 2 and 3 show the breakdown of electricity supplies in India and China in order to provide a comparison (the Appendix shows the breakdown of electricity supplies by region). Table-1: Projection for Demand and Supply
No. Year Firm Supply (MW) Peak Demand (MW) Surplus/(Deficit) (MW)
997 2.343 18.423 12.206 20.953 28.075 30.707 30.941 22.257 37.902 5.412 22.944 25.78 Azam Amjad Chaudhry Table-2: Commercial Energy Supplies Electricity Thermal Hydroelectric (Hydel) Nuclear Fiscal Year Installed Generation Installed Generation Installed Generation Imported Capacity (Gwh) b Capacity (Gwh) b Capacity (Gwh) b (Gwh) (MW) a (MW) a (MW) a 1990/91 1991/92 1992/93 1993/94 1994/95 1995/96 1996/97 1997/98 1998/99 1999/2000 2000/01 2001/02 2002/03 2003/04 2004/05 2005/06 2006/07 2007/08 Jul-March 2007/08 2008/09 2.858 23.700 12. .478 12.480 6.494 6.491 6.077 2.826 4.456 7.855 10.288 3.741 5.707 21.482 6.375 27.169 12.299 12.647 21.826 4.795 2.112 19.481 18.006 9.291 1.354 26.174 51.041 6.898 3.669 42.288 17.283 63.857 5.499 6.479 6.436 12.162 60.591 52.825 137 137 137 137 137 137 137 137 137 137 462 462 462 462 462 462 462 462 462 462 385 418 582 497 511 483 346 375 284 399 1.137 8.122 57.760 2.286 12.740 1.696 10.176 33.972 63.858 22.484 2.319 918 0 73 109 146 171 199 202 195 12.449 19.041 5.194 18.826 4.826 4.862 31.060 22.526 5.330 4.921 39.826 4.632 p 39154 p Source: Hydrocarbon Development Institute of Pakistan (HDIP).823 6.436 22.726 4.626 4.489 12.640 20.669 46.351 26.285 12.478 12.826 4.877 49.064 48.671 30.926 51.
Figure-3: China-Electricity Generation 1.A Panel Data Analysis of Electricity Demand in Pakistan 79 Figure-1: Pakistan-Electricity Generation 0.8 0.8 Proportion of Total Electricity Generated 0.6 0.8 0.6 0.4 0.2 0. .0 1975 1980 1985 1990 1995 2000 2005 2010 Year Thermal(%) Hydel(% Nuclear(% Source: International Energy Agency (2009).4 0.2 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(%) Hydel(%) Nuclear(%) Year Source: International Energy Agency (2009).0 Proportion of Electricity Generated 0.6 0.2 0.0 1975 1980 1985 1990 1995 2000 2005 2010 Year Thermal(%) Hydel(% Nuclear(% Source: International Energy Agency (2009).0 0. Figure-2: India-Electricity Generation Proportion of Electricity Generated 1.4 0.
Figures 5 and 6 show the price of electricity in the household and industrial sectors of Pakistan.000 50. prices have stabilized. based on the most common tariff categories.000 20.000 10.000 70. but rather the average price of electricity in each sector based on the tariff categories into which most household and industrial consumers fall. The rate of growth of demand for electricity in the industrial sector (despite a dip in 2005) increased after 2000 and is higher than the growth in demand in the agricultural and commercial sectors.000 30. .075 and the price of electricity in India for industries falls roughly into the range of $0. 2009).000 60.000 40. Note that this is not the average price across all tariff categories.80 Azam Amjad Chaudhry Figure-4: Energy Consumption in Pakistan 80.065-0.000 1990 1995 2000 Year 2005 2010 Household Commercial Industrial Agricultural Total Energy Consumption (Gwh) 1 Source: Pakistan Economic Survey (various years).085-0. As a point of comparison. the price of electricity has been increasing over time. while in US dollar terms. the price of electricity in India for households falls roughly into the range of $0.1 Note that in rupee terms.092 (Government of India. Figure 4 shows the history of energy consumption in Pakistan and shows the higher rate of growth in the demand for household electricity as compared to the demand for electricity in other sectors.
Note: The prices are estimated based on the tariff categories into which households (Tier 3:101-300 units) and industries (Category B2) most commonly fall. Note: Prices are estimated based on the tariff categories into which households (Tier 3:101300 units) and industries (Category B2) most commonly fall.04 0. Figures 7 an 8 shows the relationship of the average price of electricity over the period 1998–2009 and the real gross domestic product .06 0.12 0.1 0.08 $/Kwh 0. Figure-6: Price of Electricty Nased on Most Common Tariff Categories ($/KWh) 0.02 0 1985 2005 1995 2000 Year 1990 2010 Household Price (Rs/KWh) Industrial Prices (Rs/KWh) Source: Pakistan Economic Survey (various years). The exchange rate used is the average yearly exchange rate.A Panel Data Analysis of Electricity Demand in Pakistan 81 Figure-5: Price of Electricity Based on Most Common Tariff Categories (Rs/KWh) 6 5 4 Rs/KWh 3 2 1 0 1985 2005 1995 Year 2000 1990 2010 Industrial Prices (Rs/KWh) Household Price (Rs/KWh) Source: Pakistan Economic Survey (various years).
Figure-8: Average Price of Electricity for Industries Across Countries (Using Averages from 1998-2008) 0. across countries. household electricity prices rise but industrial electricity prices do not. So as real income grows in a country. .15 0.25 0. 2 Note that these are simply nominal prices taken from the International Energy Agency.25 0.1 0.2 The interesting fact that arises is that.2 0.05 0 0 2000 y = -8E-07x + 0.0994 R2 = 4000 6000 GDP Per Capita ($/KWH) 8000 INDPRICE Linear (INDPRICE) Source: International Energy Agency (2009) and Heston. but a weakly negative relationship between average industrial household electricity prices and real income.82 Azam Amjad Chaudhry (GDP) per capita for a sample of 63 countries.05 0 y = 9E-07x + 0. Figure-7: Average Price of Electricity for Household Across Countries (Using Averages from 1998-2008) 0.3 Average Price of Electricity for Households ($/KWH) 0. Summers and Aten (2009). Summers and Aten (2009).1 0. there is a weak positive relationship between average household electricity prices and real income.1074 R2 = HPRICE Linear (HPRICE) 0 10000 2000 3000 4000 50000 GDP Per Capita ($) 6000 7000 Source: International Energy Agency (2009) and Heston.15 0.2 Industries ($/KWH) 0.
A Panel Data Analysis of Electricity Demand in Pakistan 83 III. Per capita consumption of electricity is far more revealing: China and South Korea have had dramatic increases in their per capita consumption (primarily due to increases in industrial output and demand). they should be formulating a strategy to meet potential energy demand in the next 4 to 5 years. South Korea. but also with determining the potential gap between electricity supply and electricity demand in the future. Cross-Country Analysis of Demand for Electricity Economists and policymakers are concerned not only with the impact of electricity shortages on consumers and industries. Figure-9: Consumption of Electricity in a subsample of Countries (Billion KWh) 3000 2500 Consumption Electricit 2000 1500 1000 500 0 1970 1980 1990 Years 2000 2010 China India Indonesia Korea. while per capita consumption in Pakistan. India. The purpose of this section is to look at cross-country evidence to determine what happens to the total demand for electricity as per capita GDP increases. there has been a dramatic increase in total electricity consumption in China. While policymakers struggle with fulfilling the current level of demand. it would be a critical mistake to try and aim toward the target of energy consumption today. and Pakistan). Rather. while increases in energy consumption in the other four countries have occurred at a constant rate. India. the elasticity of electricity consumption with respect to per capita income is calculated. Pakistan . Figure 9 and Figure 10 show total electricity consumption and electricity consumption per capita in five countries (China. Indonesia. Using a unique panel dataset with 66 countries. As can be seen. and Indonesia have grown at about the same rate.
South Pakistan Source: International Energy Agency (2009). 3 . The fact that one has data for 66 countries across a period of 10 years3 means that one can use a simple fixed effects model to estimate the elasticity of electricity consumption per capita with respect to per capita income. countries with higher real per capita GDP had higher levels of electricity consumption per capita.84 Azam Amjad Chaudhry Figure-10: Consumption of Electricity per Capita in a Subsample of Countries (KWh) 8000 7000 6000 Consumption per 5000 4000 3000 2000 1000 0 1975 1980 1985 1990 1995 Years 2000 2005 2010 China India Indonesia Korea. one can see that for the period 1998–2008. Figure 3 shows average per capita consumption of electricity for a group of countries. It should be noted at this stage that the data from the International Energy Agency did not contain consumption and price data for all years for each country and thus the panel is unbalanced. In it.
Econometric Model The basic formulation used to look at the relationship between per capita consumption of electricity. The fixed effects (or within groups) estimator is based on the transformed equation in which the country specific effect is eliminated: (electrit-electribar) = β1(pcincomeit-pcincomeibar) + β2(prelectit-prelectibar) + (εit. Summers and Aten (2009).εibar) . real per capita income and electricity prices is as follows: electit = β1pcincomeit + β2prelectit + ηi + εit where: electit is the per capita consumption of electricity in country i at time t. prelectit is the average price of electricity in country i in time t. pcincomeit is the real per capita income of country i in time t.A Panel Data Analysis of Electricity Demand in Pakistan 85 Figure-11: Consumption Per Capita of Electricity Across Countries (Average 1998-2008) 30000 Electricity Consumption per Capita (KWH) 25000 20000 15000 10000 5000 0 0 10000 20000 30000 40000 50000 60000 70000 Real GDP Per Capita ($) CONSCA Source: International Energy Agency (2009) and Heston. and εit is the idiosyncratic error term that is uncorrelated with the explanatory variables. ηi is the time invariant country specific effect that may be interpreted as the state of technology in country i or level of natural resources.
86 Azam Amjad Chaudhry Under the assumption of exogeneity of explanatory variables. They show that the elasticity of per capita demand for electricity with respect to per capita income is approximately 0. 2009). a panel with 63 countries over a 10-year period is formed. Data on population and real per capita GDP is taken from the Penn World Tables (Heston. This means that a 5% increase in per capital income in Pakistan will be accompanied by a 3. back-of-the-envelope calculation. Summers and Aten. Description of the Data The data used is in this analysis is taken from the International Energy Agency’s (IEA) Energy Prices and Taxes. if real GDP growth is approximately 5% per year for 5 years and population growth is approximately 2% per year. 2009). Empirical Results The results of the fixed effects estimation are shown in Table-1. Excluding missing data. . they should be looking to meet both current electricity demand plus another 9%.5% increase in the demand for electricity per capita in Pakistan. then electricity demand will increase by approximately 9%. the fixed effects estimators give unbiased estimates of the elasticity of electricity consumption per capita with respect to real per capita income and electricity prices. As a simple.69. Quarterly Statistics (3rd Quarter. The IEA report contains data on economy-wide electricity consumption and average household and industrial electricity prices for 66 countries for the period 1998–2008. So if policymakers are looking at projects to meet the current energy shortfall and these projects come on line in about 5 years.
012 (. because low.and middle-income countries (based on the World Bank definition of low.and middle-income countries have a greater elasticity of electricity consumption than higher-income countries.025) 1.042** (. . The analysis above was also performed for low.297) 63 0.687** (.and middleincome countries may be in the process of rapid industrialization which could have a significant impact on their consumption per capita.026) -0. Before one moves on. **Significant at 1% level.875 2 344 Fixed Effects 0. one may want to consider the possibility that low.104) .282** (.and middle-income countries) and the results are shown below. As can be seen. the elasticity of electricity consumption with respect to income per capita is not significantly different from the estimates for the entire sample.A Panel Data Analysis of Electricity Demand in Pakistan 87 Table-1: Estimates of Per Capita Demand for Electricity for All Countries Model Log(GDP Per Capita) Log(Price) Constant Number of Countries R2 R2 within Parameters Observations Pooled Cross-Section 1.616 2 344 *Significant at 10% level.037) -5.325** (.04) 63 0.843 0.59 (1.
and businessmen have increasingly concerned themselves with the demand by the household and manufacturing sectors for electricity.65** (0.083) -. At the firm level. the availability of better quality disaggregated data has enabled researchers to begin focusing on micro-level analyses of the demand for electricity. economists.88 Azam Amjad Chaudhry Table-2: Estimates of Per Capita Demand for Electricity for Low and Middle Income Countries Model Log(GDP Per Capita) Log(Price) Constant Number of Countries R2 R2 within Parameters Observations Pooled Cross-Section 1.784 2 117 Fixed Effects 0.695 2 117 *Significant at 10% level. IV.732 0. the question of the impact of higher electricity prices on standards of living and on manufacturing output has become critical from both a welfare perspective and an economic growth perspective.993) 28 0.113) 0. Micro Level Analysis of Demand for Electricity in the Industrial Sector of Pakistan Over the last decade. As the price of electricity facing both households and firms has increased.212** (0. such as the household or manufacturing sector) to changes in the price of electricity. a few studies such as Woodland (1993) and Bjorner et al. policymakers. with an emphasis on either household demand for electricity or firm-level demand for electricity.73) 28 0. .057) -7. (2001) use panel data on firms to see the impact of changes in the .60** (0. Most analyses of the demand for electricity use time series data and macroeconometric tools to look at the response of economy-wide demand for electricity (or the demand for electricity by a particular sector. The purpose of this section is to use firm-level data to determine the elasticity of manufacturing output with respect to the price of industrial electricity.62** (0 . Over the last decade.036 (0.52 (0.033) 1. **Significant at 1% level.
Using this data. f(x1.y) / ∂pi i = 1. The main benefit of a panel data set is that it allows one to relax the strict assumptions about the uniformity of the estimated parameters. xn) ≥ 0 where (p1 + …pn) are given factor prices for the n inputs. the leather sector..y) = Min (p1x1 + …pnxn) s.pother. Thus the firm’s short-term problem of minimizing production costs can be expressed as: C(p. xn) are the quantities of of the n inputs used. and a final sector that combines the electronics and other manufacturing sectors. If one makes the assumption that firms consider the price of electricity and other factors of production to be exogenous and that each firm minimizes its cost function. The subsectors analyzed are the textile sector.y) = ∂C(p.. Shephard’s Lemma yields the factor demand functions Xi(p.. .….…. y is the level of output and f(x1. The analysis in this section uses a relatively new panel data set containing data on 402 firms over 4 different years to determine the impact of changes in the price of industrial electricity on the demand for electricity by Pakistani manufacturing firms. the demand function can be simply expressed as a function of the factors of production and the level of production (Bjorner.n So the demand for electricity can be expressed as: E = f(pe.…. electricity is used as an input in their production functions. pe is the price of electricity. xn) is the production function indicating the feasible output.….y) where E is electricity demand. the food sector.A Panel Data Analysis of Electricity Demand in Pakistan 89 prices of electricity on firm-level output. Microeconomic Foundations of Estimated Model For firms. xn) ≥ y and (x1. pother is the prices of the other factors of production and y is the level of output. 2001). a fixed effects model is estimated for all 402 firms to find the aggregate elasticity of electricity demand with respect to price. the chemical sector.y): Xi(p. This is followed by a disaggregated analysis in which the firms are divided according to subsector to determine each subsector’s specific price elasticity of demand.t. (x1.
one finds that there is very little consensus about the ‘correct’ functional form. it is much more likely that the elasticities will vary across industries because . and LPEit is the average price of electricity faced by firm i in time t. a simple log linear form is adopted. In this paper. although most studies use either a linear or a logarithmic form. represents the average variable for country i across time. temperature. sector of production. In the model above. bar.d N(0. also adjusted by a price deflator. it is assumed that all parameters are equal across firms. In actuality. Econometric Model When deciding the function form. the fixed effects estimators are derived as follows: ( LEit . Similar to the estimators from the last section.νibar ) where the subscript. it is possible to exclude the price of other inputs from the model.LEibar ) = β1 ( LFYit . It should be noted that the simple model above assumes that the parameters measuring elasticities of electricity demand with respect to output and prices are assumed to be the same across all firms.LPEibar ) + ( νit.σν2I) where LEit is the log of electricity consumption in time t for the ith firm measured in KWH.90 Azam Amjad Chaudhry If one makes the assumption that electricity is separable from other inputs so that the relationship between electricity with other inputs is neutral in terms of price. which captures all unobservable heterogeneous company-level variables that affect electricity demand. government policies.LFYibar ) + β2 ( LPEit . such as management ability.i. similar to Bjorner et al (2001): LEit = αi + β1LFYit + β2LPEit + λt + νit νit ~ i. except for αi. The parameter β1 measures the elasticity of electricity demand with respect to output and the parameter β2 measures the elasticity of electricity demand with respect to electricity price. such as technological changes. etc. etc. LFYit is the log of yearly output in rupees in time t of firm i adjusted by a price deflator for output. The parameter λt captures the effect of changes in unobserved variables that influence companies equally.
firm-level data was taken for the value of annual output and annual expenditures on electricity. and annual electricity expenditures are taken from the World Bank’s Enterprise Surveys. Description of the Data The data is taken primarily from two sources: Firm-level data on annual output. but also collected some (though not all) data for the ‘previous year’. Then for each category of industrial electricity consumers (B2. annual sales. Since the firm-level surveys only had data on annual expenditures on electricity. each of the 402 firms was placed in a tariff category. the tariff structure for industrial electricity was taken from the Economic Survey of Pakistan. and expenditures on electricity for 4 years (2001. There were a total of 402 firms for which data was collected in both surveys. After this. costs of inputs and labor. based on average annual usage of . The Enterprise Survey conducted in Pakistan gathered information on firm-level characteristics. 2002. Based on these ranges. For the purposes of the analysis in this section. in 2002 and 2006/07. B3.A Panel Data Analysis of Electricity Demand in Pakistan 91 different types of companies respond in different ways to changes in output and changes in electricity prices. and indicators of business environment. and were answered by business owners and top managers. 2005/06 and 2006/07). One could thus gather firm-level data on firm-level characteristics. and B4). focused on the manufacturing sector. the chemicals sector. The Economic Survey of Pakistan was used to calculate the average price of electricity faced by each firm. The firm-level surveys were conducted for a representative sample of firms in the private sector. Thus. a range was created into which the annual expenditure on electricity would fit. An extremely useful fact of the survey was that it not only collected data for the year of the surveys (2002 and 2006/07). For this reason. the leather sector. the model above is also estimated for the following industrial subsectors: the textile sector. 2002 and 2006/07 while data on industrial electricity tariff structures is taken from the Economic Survey of Pakistan. the tariff structure was again used to calculate an average price of electricity in Rs/kWH for each tariff category. the food sector. and a final sector that combines the electronics and other manufacturing sectors. The World Bank Enterprise Survey (conducted as a part of the World Bank’s Investment Climate Assessment for Pakistan) was administered twice in Pakistan. one was able to construct a panel dataset for 402 firms for a period of 4 years. the following procedure was adopted to calculate an ‘average price of electricity’ facing each firm: For each of the years of the survey. annual sales of firms.
92 Azam Amjad Chaudhry electricity. Also. although he chose firms that used coal. they are higher than estimates obtained from similar studies in other countries.7% decrease in the demand for electricity. Empirical Results The results for the sample of firms is given in Table-3. the demand for electricity in the food sector is more elastic with respect to price (-0.06 across Danish industrial subsectors. one obtained data on annual electricity consumption Eit (in kWH). annual output FYit (in rupees) and annual average price of electricity PEit (in Rs/kWH) for each of the 402 firms for 4 years. The only other significant result is that the price elasticity of demand for electricity in the electronics and other manufacturing sector is equal to -0. The results could imply two things: First.9% increase in the demand for electricity.57 and the output elasticity of the demand for electricity is 0. From these two sources. and LPEit. which shows both the pooled OLS estimates and the fixed effects estimates. The results for the estimation of the model for the subsamples according to sector is given in Table-4. As the results show. This average price was then used in conjunction with the firmlevel data on annual expenditures on electricity to calculate the consumption of electricity for each firm (in kWH). Also. and electricity which would allow firms to switch away for electricity as an input when faced with higher prices. In order to estimate the log-linear form of the electricity demand function (described above). if the textile and food sectors are the ones with the most private electricity generating capacity .4 to 0.67) than the average firm. The results imply that a 10% increase in the price of industrial electricity will lead to a 5. (2001) found price elasticities in the range of -0. Although the price elasticities are all less than 1.194. the price elasticity of electricity demand across the sample of firms is -0. LFYit.31. natural logs were taken of the variables to obtain LEit. Bjorner et al. gas. As the results show.81) which is significantly higher than the mean price elasticity discussed. making the demand for elasticity relatively inelastic with respect to price. the textile sector has the highest price elasticity of demand for electricity (-0. a 10% increase in firm level output will lead to a 1. Woodland (1993) found price elasticities greater than 1 for a sample of Australian firms. oil. while the other price elasticities (for the chemical and leather subsectors) are insignificant.
then they would be able to reduce their demand for external electricity the most easily. 4 .007) -0.A Panel Data Analysis of Electricity Demand in Pakistan 93 (generators). Unfortunately. this distinction was not made clear in the survey.942** (0.27 2 1467 *Significant at 10% level.4 Second.574** (0.552** (0.107) 2. then their overall electricity usage (and overall output) is most susceptible to increases in electricity prices.69** (0. if the firms in these sectors are not all generating their own electricity.5583 0. **Significant at 1% level. although switching the own-generated electricity may substantially increase their costs. Table-3: Estimates of Electricity Demand Function for All Firms Model Log(Output) Log( Price) Constant Number of Firms R2 R2 within Parameters Observations Pooled Cross-Section 0.194** (0.297** (0. It is important to understand the implications of higher industrial electricity prices in both cases. The question would be far easier to answer if the question on electricity expenditures in the Enterprise Survey had separated out expenditures on outside electricity and expenditures on own-generated electricity.5624 2 1467 Fixed Effects 0.17) 399 0.332) 399 0.086) 0.017) -0.
932* (1.615 2 643 0.671* (0.492) 176 0.524 0.383 (0.358) 2.310 2 643 Model Pooled Cross.350 2 230 0.248) 0.788** (0.522 0.215** (0.Fixed Effects Section .519) 45 0.132) 0.94 Azam Amjad Chaudhry Table-4: Estimates of Demand Function for Electricity by Industrial Subsector Industrial Sub-Sector Textiles Log(Output) Log( Price) Constant Number of Firms R2 R2 within Parameters Observations Food Log(Output) Log( Price) Constant Number of Firms R2 R2 within Parameters Observations Chemicals Log(Output) Log( Price) Constant Number of Firms R2 0.807** (0.46) 62 0.609** (0.207) 2.280** (0.17) 45 0.263) 176 0.063) 0.011) -0.609 0.021) -0.030 (0.177) 62 0.334) 2.196** (0.456 0.048) -0.457 2 230 0.220** (0.44** (0.026) -0.167) 2.018) -0.478* (1.808** (0.912** (0.327** (0.084 (0.15* (0.
038 (0.12 2 311 V.300) 83 0.162) 1.048) -0.243** (0.026) 0.283) 0. **Significant at 1% level. especially in the area of agriculture.039) -0.1313 2 159 0.394** (0. the emphasis shifted from macroeconomic analyses to microeconomic analyses.567) 25 0.366* (0.271 (0.31* (0.489 2 311 0.187 0. Machinery and other Manufacturing Log(Output) Log( Price) Constant Number of Firms R2 R2 within Parameters Observations *Significant at 10% level.95** (0.12** (0.272** (0.289 (0. In the 1960s. Micro-Level Analysis of the Production Functions of Pakistani Firms Analyses of production functions date back to the 1920s.273) 3.187 2 94 0.141** (0. when macro-level data was used to estimate an economy-wide production function in order to test marginal productivity theory and labor market competitiveness.546 2 94 0. 2 159 0.A Panel Data Analysis of Electricity Demand in Pakistan 95 R2 within Parameters Observations Leather Log(Output) Log( Price) Constant Number of Firms R2 R2 within Parameters Observations Electronics.122* (0.862) 25 0.660) 83 0. .014) -0.487 0.182) 2.
and also for firms in particular subsectors. the estimation of a firmlevel production function with electricity as an input can give us an estimate of the elasticity of output with respect to electricity. As in the analysis in the previous section. kit is the log of capital used by firm i in time t. Pakistan faces a severe shortage of electricity which is set to continue into the indefinite future. In more recent times. and a final sector that combines the electronics and other manufacturing sectors. eit is the log of electricity usage of firm i in time t. typically labor and capital. Also like the previous section. which in turn can provide us with estimates of the impact of decreases in the amount of available electricity on manufacturing sector output. the chemicals sector. This paper adopts a within-firms estimator for a Cobb-Douglas production function originally based on the work of Mundlak (1961) and Mundlach and Hoch (1965): yit = αLlit + αKkit + αEeit + ωit + εit where: yit is the log of output of firm i in time t.96 Azam Amjad Chaudhry Usual estimations of the production function focus on two factors of production. the analysis is performed for all firms in the sample. ranging from simple OLS to GMM estimation for dynamic panel data. The reason for doing this concerns the realities facing the Pakistani manufacturing sector. the food sector. the leather sector. Unlike the other economies for which firm-level production functions (and price elasticities of electricity demand) are estimated. other factors of production have been included in the estimation process. lit is the log of labor employed in firm i in time t. The analysis in this section extends the normal firm level production function to include electricity. . Econometric Model As discussed in Aguirregabiria (2009). this analysis employs the panel dataset from the World Bank’s Enterprise Survey to estimate a fixed effects model of a firm-level production function. Again. ranging from land to human capital. the subsectors analyzed are the textile sector. there are numerous ways to estimate firm-level production functions. For this reason.
ωibar) +(εit.A Panel Data Analysis of Electricity Demand in Pakistan 97 ωit = ηi + δt + ωit*. for this analysis. and annual electricity consumption. implying that a 10% decrease in electricity inputs can lead to a 5% decrease in industrial output. we can derive the fixed effects estimator of the production function which removes the fixed affect ηi: ( yit . and a final sector that combines the electronics and other manufacturing sectors. The results show that the coefficient of (log) electricity in the production function is 0. Thus. some of the data on the variables used in the estimation of the production function (labor employed and capital stocks) were not available for the year before the survey year. the model is also estimated for the textile sector. where ηi is a time invariant firm-specific effect such as the quality of a fixed input such as managerial ability. (ii) idiosyncratic shocks. annual amount of labor employed. So. Based on the assumptions that (i) idiosyncratic shocks. ωit*. although one . or land quality.libar) + αK ( kit . However. After the model is estimated for all the firms. ωit*. they need to be viewed with caution with regard to policymaking.yibar) = αL ( lit. the food sector. the chemicals sector. annual capital stocks.502. δt is an aggregate shock affecting all firms and ωit* is a firm level idiosyncratic shock. it is difficult to predict exactly what happens to output. Since we are unsure of the external/own-generated breakup of the firms electricity inputs.εibar) Description of the Data The firm-level dataset from the Enterprise Survey for Pakistan (and described in the last section) was used to gather firm-level data on annual output. While these results are significant. then output falls by 5%. are not serially correlated. the leather sector. although electricity is a critical factor of production across firms. and (iii) that the amount of inputs depends on some other exogenous time varying variables. the firm-level data for 402 firms for 2 years (2002 and 2006/07) was used. Empirical Results The estimation results for the entire sample of firms are given in Table-5. are realized after the firm decides on its level of inputs. because a 10% fall in external electricity supplies may be partially (or fully) offset by private electricity generation. unlike the dataset used in the last section. one cannot simply say that if external electricity supplies fall by 10%.kibar) + αE ( eit. and εit an idiosyncratic error term.eibar) + ( ωit .
64 Fixed Effects 0.170** (0.475) 386 0.166) 9.29) 386 0. **Significant at 1% level. Electricity shortages will either force smaller firms without own-electricity generating capacity to significantly decrease their output.597** (0. Table-5: Estimates of Production Function for All Firms Model Log(Capital) Log( Labor) Log(Electricity) Constant Number of Firms R2 R2 within Parameters 3 *Significant at 10% level. This could be due to smaller sample sizes in each analysis (because of the smaller panel used in the analysis). or will force firms with their own electricity generating capacity to face significantly higher input prices. the Pakistani manufacturing sector is heavily reliant on labor and electricity and far less on capital.108) 0.98 Azam Amjad Chaudhry can be sure of shortages in external electricity supplies and increased dependence of own-generated electricity.035) 0. and significantly higher than the elasticity of output with respect to capital. with the exception of the food subsector (which seems to be unrealistically high).636 0. Pooled Cross-Section 0.502** (0. one has to realize that with its present levels of technology. The estimations for different manufacturing subsectors are given in Table-6.057) 0.614** (1.508** (0.195** (0.914** (0.799** (0.075) 0.089) 6.257 3 . Analyzing the full sample results again show that electricity is a critical input in the production function across firms and that its elasticity of output with respect to electricity inputs is as high as that of labor. In terms of policymaking. Note that the elasticities of output with respect to electricity inputs are found to be insignificant across most industries.
554 0.428) 10.725) 169 0.031 (3.01) 0.159* (0.42) 0.299 (0.Fixed Effects Section .63) 169 0.157) 0.125) 6.518** (0.217** (0.248) 0.19** (1.03 (0.183** (0.160) 0.473 3 106 0.091) 0.150) 0.678 3 248 0.934** (1.15) 0.324** (0.109) 0.91* (1.613** (0.572 3 106 0.523** (0.082) 0.278* (0.155) 1.649** (0.205) 2.093 (0.31) 62 0.237) 10.055) 0.63** (1.261) 6.849** (0.408** (0.660 0.A Panel Data Analysis of Electricity Demand in Pakistan 99 Table-6: Estimates of Production Function for Firms According to SubSector Industrial Sub-Sector Textiles Log(Capital) Log( Labor) Log(Electricity) Constant Number of Firms R2 R2 within Parameters Observations Food Log(Capital) Log( Labor) Log(Electricity) Constant Number of Firms R2 R2 within Parameters Observations Chemicals Log(Capital) Log( Labor) Log(Electricity) Constant 0.18) 62 0.86** (0.152) 5.42** (0.243) 0.502* (0.61** (2.283 3 248 Model Pooled Cross.399* (0.082) 0.
500 0.661 43 0.533) 4.16** (1. In addition to this.25** (1.259 3 35 -0.68) 82 0.58* (2.105) 1.291) 0. Machinery and other Manufacturing Log(Capital) 0.291) 2.284 3 70 0.101** (0.98** (3. the paper looks at the impact of electricity prices and electricity inputs on manufacturing demand for electricity and manufacturing output.685 0.272 3 138 Number of Firms R2 R2 within Parameters 3 Observations 35 Electronics. VI.618 R R2 within Parameters 3 Observations 138 *Significant at 10% level. **Significant at 1% level.095** (0.128 (0.121 (0.64) 22 0.330 (0.10) Number of Firms 82 2 0.035 (0.102) 0.866** (0.445 0.442) 15.151 (0.19) 22 0. Conclusions The analyses in this paper were an initial attempt in understanding the long-term relationship between income per capita in Pakistan and the expected growth in electricity consumption per capita.12** (0.316) -0.020 (0.076) Log( Labor) 1.712** (0.100 Azam Amjad Chaudhry Number of Firms R2 R2 within Parameters Observations Leather Log(Capital) Log( Labor) Log(Electricity) Constant 43 0. .223) Constant 8.135) Log(Electricity) 0.259) -0.693 3 70 0.024 (0.915) 13.
significant impact on industrial demand for electricity and industrial output. as this paper shows. the production function estimates show that output in the manufacturing sector is highly dependent on electricity inputs. continued electricity shortages will have a significant impact on manufacturing output. as the pressure to increase electricity prices mounts. both because of significant increases in household consumption (which has already begun) and eventually higher industrial consumption. the switch to own power generation can have a significant impact on their cost of production. So. In the medium.and high-scale industries with their own electricity generation capacity. will significantly reduce output. any policies that lead to significant short-term increases in industrial electricity prices will also have to take into account the impact on manufacturing sector output. The price elasticity of demand for electricity is approximately -0. there will be strong household-level consequences (in terms of impacts on standards of living and on poverty) and.A Panel Data Analysis of Electricity Demand in Pakistan 101 The important conclusions that arise from the paper are: First. the consumption of electricity per capita will increase (at a rate of 0. Severe electricity shortages in small.65% increase in per capita consumption of electricity for every 1% increase in per capita income) at a relatively uniform rate. as income per capita in Pakistan increases.and medium-scale industries. Also. . Third. as the industrial base of Pakistan expands. Second. which do not have their own electricity generation capacity.7). This means that higher electricity prices will significantly reduce the demand for electricity in the manufacturing sector. although it is higher in the textile sector (-0.8) and in the food sector (-0.6 in the manufacturing sector.
Energy and Mining Section Board Discussion Paper. DC. Economic Survey of Pakistan. Econometric Issues and Methods in the Estimation of Production Functions. Working Paper Number 15973. (2005). V. 2003-2006).C. A Simple Method for Evaluating the Marginal Cost of Unsupplied Electricity. Federal Bureau of Statistics. M. Lessons from the Independent Private Power Experience in Pakistan. Government of India (2009).Energy Demand Models for Policy Formulation: A Comparative Study of Energy Demand Models. The World Bank Group: Washington. (2009). Government of Pakistan (Various Versions. Surrey Energy Economics Discussion Paper 108. Government of India. Development Research Group. Economic Survey 2008-2009. Munich Personal RePEc Archive. J. Federal Bureau of Statistics.. (1982).M. Dimitropoulos. T. 4866.H. 13(1). Bental.. H. AKF Forlaget. and Ravid. Methodological Advances in Energy Modelling: 19701990.R.102 Azam Amjad Chaudhry References Aguirregabiria.A. Energy Economics. Fraser. Bhattacharyya. S. and Judge.M. 249-253. Energy Journal. University of Surrey. Census of Electricity Establishments.Estimating Underlying Energy Demand Trends using UK Annual Data.. and Jensen. 111-14. Energy & Mining Statistics Section: Islamabad. J. Hunt.. The Bell Journal of Economics.H.. 14. (2000). Industrial Energy Demand and the Effect of Taxes. Copenhagen. Agreements and Subsidies. L.B. G. Bjørner. Government of Pakistan (Various Versions. and Timilsina. (2004). G. B. Griffin. 595-617. The World Bank. J. 23. 14(1). (1993). Bjørner. . Policy Research Working Paper No. (2001). T.C.. S. and Jensen. H. Togeby. 1992-2006). Industrial Companies’ Demand for Electricity: Evidence from a Micropanel. Ministry of Finance. (2009).B..
R. (2004). The Opportunity Cost of Electricity Outages and Privatization of Substations in Nepal. 815-821. Khan. K.van. . Kleijweg. M. A. Penn World Table Version 6. (eds) (2007).S. 77. B. Canada. Quarterly Statistics. (1990). 1557. Summers. Munich Personal RePEc Archive (MPRA) Paper No. Policy Research Working Paper No. 660-674.Understanding the Cost of Power Interruptions to U. Muchhala. 19781986. Usman. A. A Pooled Dynamic Analysis of Interfuel Substitution in Industrial Energy Demand by the G-7 Countries. 197688.. A. A. Heston. and Jenkins. Jyoti. The World Bank. Fueling the Future: Meeting Pakistan’s Energy Needs In The 21st Century. (2009).. C.171-181. Energy Demand in Pakistan: A Disaggregate Analysis.H. R.. (2006).T. Queen’s University. G. Energy Prices & Taxes. Ishiguro. G.. (2009). International Economics Department. International Energy Agency Publications (IEA): Paris...3. and Eto. M. S. Huigen. M. (1995). A. California 94720. Small Business Economics. and Aten. International Energy Agency (2009). and Heshmati. Electricity Demand in Asia and the Effects on Energy Supply and the Investment Environment. K..M.. 15056. 1066. 3. Ozbafli. Center for International Comparisons of Production. Jones. J. Kumbhakar. T. and Kugelman. 28. Queen’s Economics Department Working Paper No. R.H.. A.C. 2. R.. and Zeelenberg.. Journal of Applied Economics. LaCommare. Electricity Consumers. Ernest Orlando Lawrence Berkeley National Laboratory: Berkeley. Leeuwen.. Commodity Policy and Analysis Unit. Firm Size and the Demand for Energy in Dutch Manufacturing.. Woodrow Wilson International Center for Scholars: Washington. (1996). DC. and Akiyama. American Journal of Agricultural Economics. (1995).. Efficiency Measurement in Swedish Dairy Farms: An Application of Rotating Panel Data. University of California. Income and Prices at the University of Pennsylvania.A Panel Data Analysis of Electricity Demand in Pakistan 103 Hathaway. and Usman. B.
3664. A.. 43.F. 84(3). (1993). Pakistan Enterprise Survey. The Cost of Power Outages to Heterogeneous Households – An Application of the Mixed GammaLognormal Distribution. A Censored Random Coefficients Model for Pooled Survey Data with Application to the Estimation of Power Outage Costs. Washington D. Pakistan Enterprise Survey. D. The Energy Journal. Consequences of Alternative Specifications in Estimation of Cobb-Douglas Production Functions. and Scarpa. (The Impact of Business Environment and Economic Geography on Plant Level Productivity: An Analysis of Indian Industry.V. Kluwer Academic Press.. and I. (1961). Urga. Hoch (1965). 814-828. Journal of Farm Economics. Pervez. and Mengistae. A Micro-Econometric Analysis of the Industrial Demand for Energy in NSW. and Walters. 14(2). D. 25. Policy Research Working Paper No. Washington D.C. The World Bank. World Bank (2002). World Bank Group. (Compilation) (2009). Y. 57-89.. R. Woodland.. G. Empirical Production Function Free of Management Bias. (eds) Applications of Simulation Methods in Environmental and Resource Economics. 552-561. Pakistan Power Sector. Dynamic Translog and Linear Logit Models: A Factor Demand Analysis of Interfuel Substitution in US Industrial Energy Demand. 33. 44-56. Consulate General of Switzerland Pakistan: Karachi. Forthcoming in Alberini. Energy Economics. and Moeltner. The Review of Economics and Statistics. T. S.. World Bank Enterprise Analysis Unit..C. A. (2003). 1-21. C.F. World Bank Group. A.104 Azam Amjad Chaudhry Lall. World Bank Enterprise Analysis Unit. Mundlak. . Econometrica. Development Research Group. Moeltner. Y. and Layton. Mundlak. Layton. K. (2004). (2002). World Bank (2007). K.
A Panel Data Analysis of Electricity Demand in Pakistan 105 Appendix-I Breakdown of Electricity Generation by Region (Source: International Energy Agency (2009).4 0.9 0.8 0.7 0.1 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(%) Hydel(%) Nuclear(%) .7 0.8 0.1 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(%) Hydel(%) Nuclear(%) Europe – Breakdown of Electricity Generation 0.2 0.4 0.4 0.6 0. Quarterly Statistics North America – Breakdown of Electricity Generation 0.5 0.2 0.3 0.5 0.7 0.2 0.1 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(% Hydel(%) Nuclear(%) Central and South America – Breakdown of Electricity Generation 0.6 0.3 0.3 0. Energy Prices & Taxes.6 0.8 0.5 0.
7 0.1 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(%) Hydel(%) Nuclear(%) .4 0.4 0.8 0.3 0.6 0.3 0.2 1 0.9 0.2 0 1975 Thermal(%) Hydel(%) Nuclear(%) 1980 1985 1990 1995 2000 2005 2010 Africa – Breakdown of Electricity Generation 0.6 0.5 0.7 0.1 0 1975 1980 1985 1990 1995 2000 2005 2010 Thermal(%) Hydel(%) Nuclear(%) Middle East – Breakdown of Electricity Generation 1.8 0.2 0.9 0.2 0.106 Azam Amjad Chaudhry Eurasia – Breakdown of Electricity Generation 0.8 0.6 0.4 0.5 0.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.